arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2007.08880cs.LGmath.OCstat.ML
收藏

用于训练残差网络的微分对策神经优化器

A Differential Game Theoretic Neural Optimizer for Training Residual Networks

Georgia Institute of Technology(佐治亚理工学院)

Guan-Horng Liu, Tianrong Chen, Evangelos A. Theodorou

摘要:

深度神经网络(DNN)训练和最优控制理论之间的联系作为算法设计的原则性工具引起了相当大的关注。微分动态规划(DDP)神经优化器是最近提出的一种方法。尽管其经验上的成功,适用性已被限制到前馈网络和这样一个自动优化启发的框架是否可以扩展到现代架构仍然不清楚。在这项工作中,我们推导出一个通用的DDP优化器,接受剩余连接和卷积层。由此产生的最优控制表示承认一个游戏理论的角度来看,其中训练剩余网络可以被解释为合作轨迹优化的状态增强动态系统。这个博弈论DDP(GT-DDP)优化器在以前的工作中享有相同的理论联系,但生成了一个更复杂的更新规则,可以更好地利用网络传播过程中的可用信息。对图像分类数据集(例如MNIST和CIFAR 100)的评估显示,与现有方法相比,训练收敛性和方差减小方面有所改善。我们的方法突出了从架构感知优化中获得的好处。

英文摘要:

Connections between Deep Neural Networks (DNNs) training and optimal control theory has attracted considerable attention as a principled tool of algorithmic design. Differential Dynamic Programming (DDP) neural optimizer is a recently proposed method along this line. Despite its empirical success, the applicability has been limited to feedforward networks and whether such a trajectory-optimization inspired framework can be extended to modern architectures remains unclear. In this work, we derive a generalized DDP optimizer that accepts both residual connections and convolution layers. The resulting optimal control representation admits a game theoretic perspective, in which training residual networks can be interpreted as cooperative trajectory optimization on state-augmented dynamical systems. This Game Theoretic DDP (GT-DDP) optimizer enjoys the same theoretic connection in previous work, yet generates a much complex update rule that better leverages available information during network propagation. Evaluation on image classification datasets (e.g. MNIST and CIFAR100) shows an improvement in training convergence and variance reduction over existing methods. Our approach highlights the benefit gained from architecture-aware optimization.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑